About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Discovery and Design of Emerging Materials
|
Presentation Title |
Accelerating the Genetic Algorithm for Structure Prediction in 2D Materials using Machine Learning |
Author(s) |
Stephen Raymond Xie, Shreyas Honrao, Anne Marie Z. Tan, Richard G. Hennig |
On-Site Speaker (Planned) |
Stephen Raymond Xie |
Abstract Scope |
We present our machine learning approach to accelerating global structure prediction by coupling the Genetic Algorithm for Structure Prediction (GASP) to surrogate machine learning models for energy prediction. Using a small number of structurally-diverse materials generated with GASP and their formation energies from density functional theory, we train interatomic potentials using support vector regression. We show that such potentials can be used to filter low-value candidates, reducing the computational cost of the genetic algorithm by eliminating materials with a high probability of having higher energy. As more materials are generated and evaluated, their inclusion in the training data iteratively improves the accuracy of the surrogate model. We discuss the tuning of radial and angular distribution functions to encode relevant physical information into machine-readable inputs. Furthermore, we demonstrate how augmenting the training data with local energies and forces improves model performance. Finally, we apply our approach to two-dimensional group-III chalcogenide systems. |
Proceedings Inclusion? |
Planned: Supplemental Proceedings volume |